1. Field of the Invention
The present invention relates to an apparatus and method for controlling the access to a storage medium to simultaneously record/read data along a plurality of channels.
2. Description of the Related Art
With remarkable progress in moving picture coding/decoding LSI (large scale integration) for microcomputers, MPEG 2 (Moving Picture Experts Group Phase 2), etc., the technology of digitizing pictures has been outstandingly improved. As a result, television broadcasting is quickly changing from an analog system to a digital system.
In satellite broadcasting using a BS (broadcasting satellite), a CS (communications satellite), etc., programs of several hundred channels are prepared in response to various requests from users.
With the growing tendency towards digital broadcasting and an increasing number of channels, and with the development of inexpensive set top boxes (STB) and digital TV (television) programs, etc., users are receiving a large volume of digital picture data at home. Therefore, it is predicted that there are an increasing number of opportunities to store data in a storage device.
To mainly aim at accumulating a large volume of picture data, a large capacity storage device such as a hard disk, an optical disk, etc. has been developed. For example, when the MPEG 2 is used, two-hour picture data can be compressed into several GB and stored.
For the above mentioned storage device, in addition to the throughput required for an input/output system of a computer, real-time processibility is newly required. In the situation, the most promising interface for realizing a home network connecting the STB, the storage device, a printer, a display device such as a TV, etc. can be a high-speed serial interface referred to as IEEE (Institute of Electrical and Electronic Engineers) 1394.
In the IEEE 1394, data such as voice, pictures, etc. are transferred in real time. Therefore, it has a unique transfer mode referred to as isochronous transfer to guarantee the transfer of data at a constant transfer rate. Therefore, when picture data is stored in the storage device, or when picture data is read from the storage device, the request for a real-time process should be satisfied based on the isochronous transfer.
In addition, with an increasing number of multi-channel systems, a process of simultaneously (to be exact, in time series) fetching data along a plurality of channels is to be performed. Therefore, it is important to check how many channels of data can be simultaneously and efficiently stored/read with the request for the real-time process satisfied.
However, in the case of the storage device, there are a number of time factors such as, in addition to a data transfer, a wait for seek, a wait for turn, verification, retry, etc. A wait for seek indicates the time required for a disk head to move to a desired track. A wait for turn indicates a time required to turn a disk until the head of a desired sector appears below the disk head. Verification indicates a process of confirming written data. Retry indicates accessing again when access first fails.
The above mentioned time factors often prevent a real time process of successfully keeping a schedule after terminating a process within a predetermined deadline. Therefore to efficiently access a disk, the order of reading/writing data, and the areas of storing and reading data have been controlled by disk scheduling according to the conventional technology.
The following methods are listed and the conventional scheduling algorithm (A. L. N. Reddy and J. C. Wyllie, “I/O Issues In a Multimedia System”, Computer 27, Mar, pp. 69–74 1994.).
(1) EDF (Earliest Deadline First): The method of prioritizing the process having the earliest deadline.
(2) LSTF (Least Slack Time First): The method of prioritizing the process having the shortest time allowance.
(3) SSTF (Shortest Seek Time First): The method of prioritizing the process having the shortest seek time.
(4) SCAN: The method of prioritizing the process having the shortest seek time in the same seek direction.
(5) SCAN-EDF: The method of first prioritizing the process having the earliest deadline, and adopting SCAN for the processes having the same deadline.
In the above listed algorithms, (1) and (2) consider the time elements only, and does not take the efficiency of disk access (shortening the seek time) into account. On the other hand, (3) and (4) consider the elements of efficiency, but does not take the elements of time into account. Therefore, they are not appropriate for a real time process. As a result, the algorithm (5) is commonly used at present because it takes both time and efficiency in disk access into account.
FIG. 1 shows the concept of the disk scheduling by the above mentioned EDF and SCAN. In this example, it is assumed that the picture data along four channels flow in a time division system. Ch1, Ch2, and Ch3 in the four channels CH1, Ch2, Ch3, and Ch4 are used in writing transmitted picture data to a disk 1, and Ch4 is used in reading picture data from the disk 1.
The conventional scheduling has been used for playback in most cases, and the simultaneous write process is not taken into account. Therefore, the picture data of each channel is collectively and sequentially stored for an easier read operation, and the address points of channels are often assigned separately. In this example, the address points indicating the write/read positions of the data of each channel on the disk 1 are separately assigned to Ch1, Ch3, Ch4, and Ch2 in this order from outer track to inner track as shown on the right in FIG. 1.
When a write process is performed, the received picture data is temporarily stored in one buffer of a double buffer 2. Then, the process of writing the first stored picture data has to be completed while the next picture data is being stored in the other buffer of the double buffer 2. For example, through Ch1, data W12 has to be stored in the double buffer 2, and the data W11 has to be written from the double buffer 2 to the disk 1 in the round T. The same process is performed for Ch2 and Ch3.
When a read process is performed, the picture data is first read from the disk 1, and temporarily stored in one buffer of the double buffer 2. While the stored picture data are being transmitted, the next picture data is to be read in advance from the disk 1, and the process of storing the read data in the other buffer of the double buffer 2 is to be completed. For example, through Ch4, while data R42 is being transmitted from the double buffer 2, the next data R43 is to be completely read from the disk 1 to the double buffer 2 in the round T.
In FIG. 1, since disk access is requested in the order of Ch1, Ch2, Ch3, and Ch4 in time series, the deadline of the process is set in the same order. Therefore, when the EDF is adopted, the disk 1 is accessed in the order of Ch1, Ch2, Ch3, Ch4, Ch1, Ch2, Ch3, Ch4, . . . .
However, since the address points of the channels Ch1 through Ch4 are arranged in a different order, the seek distance between channels is long, and it takes a long time to move the head. Especially, the address points of Ch1 and Ch2 are set apart, and it takes a longer time to write data through Ch1 and Ch2.
When the SCAN is adopted, data is accessed in the order of Ch1, Ch3, Ch4, and Ch2 sequentially from the nearest position from the head regardless of the order of access requests. After accessing data through Ch2, the seek order is inverted, that is, data is accessed in the order of Ch2, Ch4, Ch3, Ch1. In this case, in response to access requests, the write operation through Ch1 is perform first in a round, but last in the next round. Therefore, there can be the possibility that an access request is not satisfied due to a long access interval. In such a case, the double buffer 2 has to be large enough to satisfy an access request.
On the other hand, the SCAN-EDF of the above mentioned method (5) is obtained by combining the EDF and the SCAN, and can realize the scheduling with both seek time and access request order taken into account.
However, the above mentioned conventional disk scheduling has the following problems.
In the conventional scheduling, the storage device receives and transmits picture data at a fixed rate. For example, FIG. 2 shows the process of inputting picture data of six channels Ch1 through Ch6 in the storage device at a predetermined bit rate. One transponder corresponds to the capacity of one line in satellite broadcasting.
On this assumption, it is considered that the deadline of writing/reading data through channels Ch1 through Ch6 is periodically set, and the deadline is set according to the initially determined round information.
However, the picture data displayed on a digital broadcast, etc. is statistically multiplexed as shown in FIG. 3. Therefore, the transfer rate is not always constant. In this case, the rate for one transponder is constant, but the transfer rate of coded data of the MPEP 2 of each channel depends on the speed of the movement of images, thereby realizing efficient broadcasting.
When a packet is transferred in the isochronous transfer system of the IEEE 1394, data to be transferred is normally contained in the packet. However, the picture data transferred at the above mentioned variable rate, the transfer time guarantee is maintained by transmitting a dummy packet containing no data when data to be transferred is not regularly arranged due to the variable rate (in accordance with IEC (International Electrotechnical Commission) 18663 and IEEE 1394–1995).
FIG. 4 shows the procedure of the above mentioned packet transfer system. In this procedure, a 4-byte time stamp T is added to a 188-byte transport packet 3 to generate a 192 byte packet 3a. The packet is divided into 24-byte data blocks. The four (which can be any other integer) data blocks are collected as one data block packet 4, and transferred as a isochronous transfer packet.
A header H of IEEE 1394 and a CIP (Common Isochronous Packet) header for multimedia data are added to the data block packet 4. The data block dividing method is defined in the CIP header, and a reception node can re-design the transport packet 3 according to the information.
One cycle start packet S and one data block packet 4 are transferred every 125 μs. If there is no data block packet 4, a dummy packet 5 containing only a CIP header is transferred as a isochronous transfer packet.
When the conventional deadline is set under the above mentioned situation, the deadline is defined for a case severer than a normal case based on the maximum transfer rate at which all isochronous transfer packets contain the data block packet 4. Therefore, data cannot be processed thorough a larger number of channels.
Furthermore, the above mentioned scheduling methods (1) through (5) are mainly used for playback, do not prescribe a write area on a disk, and assume that pieces of the picture data are separately stored on the disk. Therefore, when data is processed through a large number of channels, it may undesirably take a long seek time.
In addition, the current disk has a large storage capacity by dividing a high-density disk into a plurality of zones (tracks), and adopting the ZCAV (Zone Constant Angular Velocity) as a disk turn control system, thereby setting a higher transfer rate for an outer zone than an inner zone. The transfer rate of the innermost zone is, for example, about 60% of the transfer rate of the outermost zone.
However, the conventional scheduling method does not consider the above mentioned plurality of zones, but assumes that data are uniformly stored in inner and outer zones on the disk, and the amount of data and the transfer rate are uniform in the inner and the outer zones. Therefore, the performance becomes worse when data at a high transfer rate is collectively written to an inner zone. Therefore, data cannot be processed through a large number of channels in this method.
Recently, an ASMO (Advanced Storage Management Optical disc) has been studied as an optical disk mainly for storage of picture data. The ASMO is a magneto-optical disk in a magnetic field modulation system, and has the capacity of 6.1 GB at maximum per side of a 120-diameter disk.
FIG. 5 shows the configuration of the ASMO.
As shown in FIG. 5, the ASMO realizes a large capacity in a land-groove storage system for storing data with high density on both land 11 and groove 12. The pitch of the land 11 and the groove 12 is set to 0.6 micrometer. The thickness of a data recording area 13 of a disk is 0.6 mm. In this example, the data recording area 13 is divided into 22 physical zones. That is, one disk has 22 physical zones. One physical zone has several thousands of tracks. Tracks are spirally formed on a disk.
Furthermore, to have a system (chukking mechanism) of setting a disk on a drive device commonly used for a CD and a DVD, the thickness of a central portion (cramp portion) 18 on which no data is recorded on the disk is set to 1.2 mm.
In each physical zone, a predetermined number of tracks are radiantly provided. Each track is divided into one or more frames 14. The frame 14 is divided into a plurality of segments. Normally, the leading segment is an address segment (ADRS) 15, and other segments re referred to as data segments 16. Clock marks 17 indicated by Δ shown in FIG. 5 are assigned to the address segment 15 and the data segment 16.
For example, 2 KB of data and ECC (Error Correcting Code) are recorded on all data segments 16 of one frame 14. Furthermore, address information, a tilt pattern, a preamble, a reserve, etc. are recorded on the address segment 15. On the address segment 15, the information is recorded as wobble on one side only in two sides forming the groove 12, that is, in a single-side wobble method. A single sided wobbled address 19 is a single-side wobble address (two-side address can be realized) indicating the position of data. In the ASMO, the number of frames per disk turn is 16 through 73.
For the land 11 and the groove 12, the pit length is 0.235 micrometer.
As described above, the ASMO is physically divided into 22 zones, but these physical zones are divided into 714 logical zones (N through M+2) from the outermost zone to the innermost zone as shown in FIG. 6. In FIG. 6, the change in the amount of data in the buffer with the elapse of time is added to the right of each of the logical zones N through M+2. The example shown in FIG. 6 indicates the process of accessing the logical zones N through N+1 of the ASMO, reading the data from the logical zones N through N+1 to the buffer, and then reading the data from the innermost logical zone M+2 to the buffer. In this case, while the head are seeking through the outer logical zone N+1 through the inner logical zone M+2, the data stored in the buffer is transferred to an external device, and the amount of data in the buffer is gradually reduced.
The time taken for accessing the logical zone M+2 after accessing the logical zone N+1 has to be within 1 second. Within 1 second, data of 1 MB is transferred from the buffer to the external device.
FIG. 7 shows the data structure of the logical zones of the ASMO. As shown in FIG. 7, in the ASMO, a logic zone 20 has a storage capacity of 8 MB, and comprises 4 MB of land 11 and 4 MB of adjacent groove 12.
The logic zone 20 is divided into a user area (hatched portion) and a spare portion (black portion). Normally, data is sequentially recorded in order from the first piece of data in a sector unit. At this time, if there is no defective sector in the user area, the data is stored only in the user area. However, if there are defective sectors in the user area, then the data to be recorded in the defective sectors is recorded in the spare area. Thus, the spare area is used as a replacing spare sector when there is a defective sector in the user area. To compensate for the above mentioned defective sector, slipping replacement (SR) and linear replacement (LR) are used. The SR is a method of sequentially recording data in the subsequent sectors. In this case, the spare area stores data with the data to be stored in the number of defective sectors shifted backwards. On the other hand, the LR is a method of replacing the defective sectors with the spare area.
FIG. 7 shows the type of SR and LR methods. (1) shows the SR method, and (2) through (4) show the LR method. There are three types of LR methods. That is,
(2) is a method of replacing a defective sector with a spare area in the same logical zone;
(3) is a method of replacing a defective sector with a spare area in the previous logical zone; and
(4) is a method of replacing a defective sector with a spare area in the adjacent logical zone.
Thus, the ASMO is designed to shortening the access time by providing a replacing area (spare area) in the grooves and the lands of the logical zones although a data replacing process is required. Several MBs are used for physical zones and about 8 MB are used for logical zones. One physical zone contains 30 through 50 logical zones.
The range of the logical zones designed based on the range (about 200 tracks) of beam jump only by the beam deflection scanning (optical seek) by the drive of an objective 30 without a seeking operation performed by moving the body of the head as shown in FIG. 8. FIG. 8 shows an accessible area (200 tracks for the maximum width from a defective block 31 to a spare block 32) by the objective 30. The speed of the optical seek by the objective 30 is approximately 5 ms at maximum. When 200 tracks are scanned in the seeking operation by moving the body of the head as in the above mentioned optical seeking operation, a double or higher speed is required.
In the ASMO, the logical zones are sequentially accessed in principle. Although the head moves by the maximum seek distance (from the innermost zone to the outermost zone) in the next step, seamlessly fetching and reading voice and picture data can be guaranteed. The access wait time taken by the movement for the above mentioned maximum seek distance is one second. Therefore, to fetch/read voice and picture data in real time in one second, an internal buffer for storing data of 1 MB.
FIGS. 9 and 10 show the concept of seamlessly fetching and reading voice and picture data in the ASMO. FIG. 9 shows the operation of writing buffered voice and picture data to the ASMO in the logical zones N through N+2. In addition, FIG. 10 shows the operation of reading voice and picture data from the logical zones N through N+2 of the ASMO to the buffer.
In the ASMO, recording and reading data through only one channel is considered, but simultaneous recording, simultaneous reading, and time-shift reading (reading while recording) are not taken into account.
The current 3.5 inch MO (Magneto-optical disk) is based on the SCAV (Zone Constant Angular Velocity) as a turn control system. Therefore, the seek time is shorter than in the ASMO. However, in the case of the 3.5 inch MO, the transfer speed is lower for inner zones than outer zones (the transfer speed for inner zones is about 60% of the transfer speed for outer zones). When the simultaneous recording and the simultaneous reading are performed on multi-channel picture data, higher performance is required, and the data cannot be correctly processed in the inner zones. The problem occurs in the HDD (Hard Disk Drive).
In the case of the ASMO, the ZCLV (Zone Constant Linear Velocity) is adopted for constant transfer speed in the entire process by changing the number of revolutions at the inner and outer zones as a turn control system. As a result, the problem with the 3.5 inch MO does not occur. However, when access is gained for plural zones, a control time to change the number of revolutions is required, which reduces the performance.
The first object of the present invention is to provide an access control apparatus and method for efficiently processing data through a larger number of channels when access to a storage medium can be performed in real time while data through a plurality of channels are being recorded and read. The second object of the present invention is to simultaneously record and read data using a storage medium storing on both land and groove.